Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.


  1. Advanced Patent Search
Publication numberUS20040148390 A1
Publication typeApplication
Application numberUS 10/665,738
Publication dateJul 29, 2004
Filing dateSep 19, 2003
Priority dateSep 19, 2002
Also published asUS7412703, US8205210, US20080301226
Publication number10665738, 665738, US 2004/0148390 A1, US 2004/148390 A1, US 20040148390 A1, US 20040148390A1, US 2004148390 A1, US 2004148390A1, US-A1-20040148390, US-A1-2004148390, US2004/0148390A1, US2004/148390A1, US20040148390 A1, US20040148390A1, US2004148390 A1, US2004148390A1
InventorsGeoffrey Cleary, Joseph Brown
Original AssigneeCleary Geoffrey Alan, Brown Joseph I.
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Low cost, highly accurate video server bit-rate compensation
US 20040148390 A1
A method, apparatus and system for adaptably distributing video server processes among processing elements within a video server such that video server operation may be adapted in a manner facilitating rigorous timing constraints.
Previous page
Next page
1. A method, comprising:
associating each of a plurality of processing elements with at least one respective video server process;
assigning priority to said processing elements according to a hierarchy of video server processes, each of said video server processes having a relative priority level with respect to other video server processes;
adjusting said hierarchy of video server processes according to at least one of monitored timing parameters, changes in system loading conditions, changes in operating conditions and operating system scheduler requirements.
2. The method of claim 1, further comprising:
adjusting said associations of processing elements and video server processes in response to at least one of processing element loading, monitored timing parameters, changes in system loading conditions, changes in system operating conditions and operating system scheduler requirements.
  • [0001]
    This application claims benefit of United States provisional patent applications Serial No. 60/412,038 and 60/413,197, filed respectively on Sep. 19, 2002 and Sep. 24, 2002, which are herein incorporated by reference in their entireties.
  • [0002]
    A key feature of a video server is its ability to generate streams of video at a precise bit rate. This is critical for compliance with the MPEG standard. A secondary goal for a video server is of-course, a low cost of its components. There is a conflict between these two goals, since cheaper components tend to have higher drift. This is particularly true of components such as crystals. Earlier video servers that do not utilize this invention required expensive, highly stable crystal oscillators to be included in their design.
  • [0003]
    Commonly available General Purpose Operating System software does not provide deterministic response times to events (such as the availability of data from a disk, or the need to output data at a given time). To solve these problems, two solutions are typically employed either together or singly. First, operating system software with deterministic behavior is utilized. This software has a premium cost associated with them in the form of runtime royalties and development seat costs. Second, custom hardware is designed to tackle the timing problems and the Operating system read/writes data to/from the hardware via a buffering system, separating the Operating system from the time critical constraints. Such hardware development and fabrication have costs and risks associated with them.
  • [0004]
    Various problems of the prior art are addressed by the present invention of a method, apparatus and system for adaptably distributing video server processes among processing elements within a video server such that video server operation may be adapted in a manner facilitating rigorous timing constraints.
  • [0005]
    Specifically, a method according to one embodiment of the invention comprises associating each of a plurality of processing elements with at least one respective video server process; assigning priority to said processing elements according to a hierarchy of video server processes, each of said video server processes having a relative priority level with respect to other video server processes; adjusting the hierarchy of video server processes according to at least one of monitored timing parameters, changes in system loading conditions, changes in operating conditions and operating system scheduler requirements.
  • [0006]
    So that the manner in which the above recited features of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to embodiments, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.
  • [0007]
    [0007]FIGS. 1 and 2 depict, respectively, structure architectures for multi-processor and uni-processor implementations according to the present invention;
  • [0008]
    FIGS. 3-10 and 12 graphically depict timing, temperature drift and other parameters useful in understanding the present invention;
  • [0009]
    [0009]FIG. 11 depicts an exemplary rate control loop suitable for use in an embodiment of the present invention;
  • [0010]
    [0010]FIG. 12 depicts a method according to an embodiment of the present invention;
  • [0011]
    [0011]FIG. 13 depicts a high level block diagram providing a plurality of actions in accordance with an embodiment of the present invention; and
  • [0012]
    [0012]FIG. 14 depicts a controller suitable for use with the present invention.
  • [0013]
    One feature of the invention is that by careful partitioning of the processes which constitute a video server system onto the individual processors of a Symmetric Multi-Processing system, the need for specialized hardware or Operating System Software is mitigated.
  • [0014]
    Initial tests, utilizing a modified Linux Operating System, have shown satisfactory results for the present invention, which will allow a substantial reduction in cost of a video server product line. One implementation uses multiprocessor version of Linux Operating System. It has been enhanced to allow control over which processes run on which processors. Another implementation is used if the first implementation proves to provide insufficient performance.
  • [0015]
    [0015]FIG. 1 shows the software architecture for a Linux implementation of the present invention (other operating systems may be used). The figure shows the foundation hardware to be Intel MultiProcessor Specification compliant hardware, running an SMP Linux Operating System. The kernel is modified to allow CPU binding (processes can be locked onto specific processors). The first processor runs the administrative processes and the video data retrieval process. The second processor runs software that drives the video data out at a critically controlled rate. Since the data streaming has the highest real time demands, interruptions of this process can be minimized by assigning all interrupts (other than Gigabit Ethernet) to the other processor.
  • [0016]
    [0016]FIG. 2 shows the software architecture for the uni-processor Linux implementation. Again the foundation hardware is Intel Multi-Processor Specification compliant hardware. The video server tasks are divided up the same way; the first processor runs the administrative processes and the video data retrieval process and the second processor runs software that drives the video data out at a critically controlled rate. The difference being that the second processor is not running Linux and consequently the application will not be interrupted by the operating system. This allows the dedicated streaming engine software to have fine control over the CPU1 cache hardware and to allow the Stream Engine to achieve maximum performance by executing it's code entirely from within the processors L1+2 cache.
  • [0017]
    Since the Streaming Engine is running without an OS support, the Kernel's IP stack will no longer take care of network housekeeping (ICMP). To facilitate this, it is proposed that a ‘Gigabit Ethernet virtual driver’ be implemented under the Linux kernel on CPU0. This would pass any non-streaming packets back and forth between the Streaming Engine code running on CPU1 and the kernel on CPU0. This allows ping and other ICMP messages to be supported without the need to implement them in the Streaming Engine code.
  • [0018]
    Intel has published a standard for MultiProcessor systems known as “MultiProcessor Specification” version 1.4 1997, which is incorporated herein by reference in its entirety and may be accessed at 1606.pdf. This specification describes an architecture where memory and peripherals are symmetrically shared between multiple processors. Machines are readily available, which conform to this standard, in the form of multi-processor PCs. The intention or the MultiProcessor Standard is for a single copy of the operating system and applications to run on all/any processors in the system. This architecture is known as Symmetric MultiProcessing.
  • [0019]
    When an application process runs on a MultiProcessor machine under a General Purpose Operating System, such as Linux, it will typically utilize the processor with the least load. The operating system has a component known as the scheduler, which periodically re-schedules tasks. It is critical that the operating system provide a mechanism to assign priorities to processes, however, even the highest priority process can be ‘swapped’ from one processor to another many times per second. Such swapping causes unacceptable inefficient utilization of the processors.
  • [0020]
    Video Servers are required to output data with strict conformance to industry standards. These standards require data packets to be output with deadlines expressed in microseconds. The demands of the output stage, in turn place requirements on the retrieval of data from the disk drives. Such requirements require the software to be Asymmetric, i.e. dedicated software runs on dedicated processors. By careful design and consideration of these needs, this patent describes how a COTS PC and COTS software can be utilized to create a platform that provides real time performance required for a Video on demand server.
  • [0021]
    The benefits of this invention are that dramatic savings can be made on the cost of the computer hardware and Operating System software. The trade off is that highly optimized and dedicated application software needs to be written.
  • [0022]
    This invention in one embodiment incorporates use of NTP; Network Time Protocol (STD0012, rfc1305 by David L. Mills, University of Delaware, March 1992). It also enhances the claims made in a related patent disclosure, “Video Streaming Server Utilizing Asymmetric architecture on a Symmetric Multiprocessing platform”. NTP provides the mechanisms to synchronize time and coordinate time distribution in a large, diverse internet operating at rates from mundane to lightwave. It uses a returnable-time design in which a distributed subnet of time servers operating in a self-organizing, hierarchical-master-slave configuration synchronizes local clocks within the subnet and to national time standards via wire or radio. The servers can also redistribute reference time via local routing algorithms and time daemons.
  • [0023]
    By use of this protocol in the design of a video server, relatively inexpensive hardware (e.g., off-the-shelf Personal Computer systems), can be used in the construction of high quality systems.
  • [0024]
    Terms and Abbreviations
    TSC Time Stamp Counter
    DSM DIVA Services Manager: A components of the DIVA Video On
    Demand System.
    NTP Network Time Protocol: rtc2030
  • [0025]
    DVS6000 Embodiment
  • [0026]
    In one embodiment, techniques are provided that will be used to achieve satisfactory performance of the DVS6000 server in terms of its accuracy of output bit-rate. The techniques described in this document address the need to be able to measure and correctly compensate for, the actual frequency and drift of the main clock crystal on the system motherboard, which is not stable or well defined.
  • [0027]
    The highest resolution clock in a typical DVS6000 system is the processor clock. This runs at approximately 1 GHz. It is derived by multiplying up a 14.31818 MHz clock. This is the same crystal that is divided down to produce the 100 Hz system clock interrupt. The division is achieved by dividing the 14 MHz crystal by 12 to get 1.1931817 MHz, and dividing this by 11932. This corresponds to an error of about 200 ppm. To correct errors in the local system clock, the NTP software is proposed. NTP is already in place to keep the DSM time synchronized to a reference. Using an accurate system clock, the GHz clock can be measured and used as a basis for timing the transmission of MPEG packets
  • [0028]
    The output bit-rate of the DVS6000 is held to within 6 ppm in order to meet the buffer requirements of the MPEG sink. The limiting factor is buffer overflow. Bit-rate of 15 is required for HDTV, we are concerned, only with bit-rates up to 4 Mbps. For a 2-hour movie, we need a clock reference accurate to within 6.37 ppm. See Appendix A. This means that the typical output can range from 3.374988 to 3.750012 Mbps.
  • [0029]
    NTP Performance
  • [0030]
    Rfc-2030 (NTP version 4) states “With careful design and control of the various latencies in the system, which is practical in a dedicated design, it is possible to deliver time accurate to the order of microseconds”. In practice, offsets between server and client of within 5 ms are achieved. FIGS. 3, 4 and 5 show the offset (in seconds) between, respectively, the client and reference server, the frequency compensation being applied to the local clock in ppm, and the estimated error of the measurement of the reference time (in seconds).
  • [0031]
    These measurements were taken using a Qsol machine to confirm that the error will be within 5 ms. The time servers we were referencing have three switches and a router between them. The original reference is a Stratum 1 Internet source. It is imperative that the quality of the original time source be good and that the network between the servers be good. Since NTP uses round trip UDP packets to acquire time samples, poor network communications increases the error in the time correction. It is assumed that DIVA will access to a stratum 1 or, optionally, stratum 2 timeserver with guaranteed sub millisecond to millisecond accuracy is provided.
  • [0032]
    The following results demonstrate the direct relationship between error reported by NTP and wander in the derived frequency of the GHz crystal.
  • [0033]
    [0033]FIG. 6 shows the calculated clock frequency over a period of 6 hours. The measurements were taken every 1000 seconds. FIG. 7 shows the error reported by NTPd over the same period. The correlation is clear. This implies that the drift of the crystal is insignificant relative to the error from the timeserver used during this test.
  • [0034]
    Should NTP be incapable of acquiring time readings from its timeservers, NTP will continue to correct the local time using the last calculated drift. The DVS6000 system should report the failure of a timeserver to respond if this should continue for more than 8 hours. See FIG. 8 that shows the behavior of NTP after the server connection is removed, over a 24-hour period.
  • [0035]
    Local Clock Stability
  • [0036]
    The stability of the local clock is based on the characteristics of the 14 MHz crystal. See FIG. 9 for its drift with temperature. Each line corresponds to a particular angle of cut of the crystal. Wee assume worst case.
  • [0037]
    In summary the 14 MHz crystal will change by 1 ppm for each degree C. change in temperature. This will result in a change of 1 ppm per degree C. in the GHZ CLOCK. FIG. 10 shows actual results measured on a qsol machine. They show a 35 degree C. change in external temperature caused a 10 ppm change in crystal frequency. This indicates that the crystal is somewhat insulated from external temperature changes. Also indicated in these results is that variations in the 120 VAC supply do not significantly affect the crystal frequency.
  • [0038]
    Rfc-1305 states “The 32-bit Skew-Compensation register is used to trim the oscillator frequency by adding small phase increments at periodic adjustment intervals and can compensate for frequency errors as much as 0.01% or 100 ppm.”
  • [0039]
    Media Server Rate Controller
  • [0040]
    The Media Server utilizes the on-chip Time Stamp Counter (which has nS precision) to transmit the frames of MPEG data at the appropriate time. Given a precise knowledge of the frequency of the TSC, the correct period between frames (Ticks per Frame) can be deduced as follows:
  • [0041]
    Frames_Per Second=MPEG_Bit_Rate/Bytes_Per_Packet/Packets_Per_Frame/Bits_Per_Byte
  • [0042]
  • [0043]
    In order to measure Ticks_Per_Second, it is necessary to count the ticks of the TSC over an accurately measured period of time.
  • [0044]
    In FIG. 11 you can see the point at which NTP corrects the system clock. These corrections do not affect the Pentium's Time Stamp Counter (TSC) which is directly driven from the crystal, consequently, the Media Server Rate Controller will continuously measure the TSC frequency and correct the value of Ticks per Second used to correctly space the transmitted frames.
  • [0045]
    The Media Server attaches an expected transmission time to each frame that is queued up and ready to transmit. The difference between consecutive frames is the value Ticks_Per_Frame. The Media Server's rate generator monitors the TSC and tries to transmit the packets at the correct time.
  • [0046]
    Initial System Calibration
  • [0047]
    Since it is important to have an accurate measurement of the TSC frequency before the server can deliver conformant streams, in one embodiment the TSC be calibrated during the factory testing. The initial calibration would be over a one-hour period. I would be essential to ensure that a suitable timeserver be available during this initial calibration. This optionally consists of a T1 connection to the Public Internet timeservers. The calibration value is held on the disk in the writable partition.
  • [0048]
    Boot Time
  • [0049]
    During boot up, the Media Server reads the last calculated value for the TSC frequency. During the next hour, the server calculates a new value; this may vary from the current value due to NTP errors and temperature variations etc. The maximum temperature change possible in a head end is 30 degrees C. This causes a 10-ppm change in frequency. In one embodiment, the maximum NTP error tolerated in crystal frequency measurement is 10 ppm. If the new value differs by more 3 than 25 ppm, this value is capped at 25 ppm and a new value calculated. Should 3 successive values be capped, a warning is generated.
  • [0050]
    Normal Operation
  • [0051]
    During normal operation, the temperature will be stable relative to the one-hour measurement period. Again, if the new value differs by more than 25 ppm, this value is capped and a new value calculated. Should 3 successive values be capped, a warning is generated. Rather than just accepting the new value, it is important to smooth out the affect of NTP error in the time measurements.
  • [0052]
    The new value is used to update the current value using a low pass filter:
  • [0053]
    The desired value of k in the equation can be estimated from the typical sample data in FIG. 10.
  • [0054]
    The solid line is the actual crystal frequencies measured at 3600-second intervals. The dotted line is a moving average. The change between points 1,130,412,954 and 1,130,412,954 is 1274 or 1.274 ppm. The actual amount we would like to change is indicated by the change in the moving average line at this point. The value of k should be 0.25. This value should be confirmed by experimentation.
  • [0055]
    The old value must be stored in a non-volatile memory such that it can be ready after a re-boot of the system. It is proposed that a region of disk be reserved for such storage.
  • [0056]
    Appendix A
    VBV Drain Calculation
    VBV size (bits) 1835008
    Effective buffer size 90%
    Buffer Headroom 10%
  • [0057]
    What output Video rate (Mbps)
    rate drains 3.0 error 3.18 error 4 error 15 error
    the buffer in: Δ bps Mbps (ppm) Mbps (ppm) Mbps (ppm) Mbps (ppm)
    1 hour 458.75 2.99954 152.92 3.17954 144.26 3.99954 114.69 14.99954 30.58
    2 hours 229.38 2.99977 76.46 3.17977 72.13 3.99977 57.34 14.99977 15.29
  • [0058]
    What output
    overflows error error error error
    the buffer in: Δ bps Mbps (ppm) Mbps (ppm) Mbps (ppm) Mbps (ppm)
      1 hour 50.97 3.00005 16.99 3.18005 16.03 4.00005 12.74 15.00005 3.40
    1.5 hours 33.98 3.00003 11.33 3.18003 10.69 4.00003 8.50 15.00003 2.27
      2 hours 25.49 3.00003 8.50 3.18003 8.01 4.00003 6.37 15.00003 1.70
  • [0059]
    Typical PC Clock Crystal
    Frequency 32.768 kHz
    Freq tolerance @ 25 C. +/− 20 ppm
    Aging 1st yr @ 25 C. +/− 3 ppm
    Freq stability −0.04 ppm/(Delta C.)2
    Temp operating range −10 to +60 C.
  • [0060]
    Typical PC Ref Clock Chip
    Frequency 14.31818 MHz
    Freq tolerance @ 25 C. TBD
    Aging 1st yr @ 25 C. TBD
    Freq stability TBD
    Temp operating range TBD
  • [0061]
    The error with the NTP corrected system clock typically is less than 5 ms. Assuming this, we will have no more than 10 ms of error when measuring a duration of time (two measurements). If we measure the GHz clock over one hour we will have up to 10 ms of error in 3600 seconds or 2.7 ppm error. The output of the server is directly driven by the GHz clock and will be 2.7 ppm.
  • [0062]
    [0062]FIG. 14 depicts a high-level block diagram of a controller suitable for use with the present invention. Specifically, the controller 1400 of FIG. 14 may be employed to implement the various software functions and/or method steps described herein, as well as control server and transmission hardware and generally implement the architectures discussed below with respect to FIGS. 1, 2 and 11. The controller 1400 of FIG. 14 comprises a processor 1430 as well as memory 1440 for storing various control programs and other programs 1442. The processor 1430 cooperates with conventional support circuitry 1420 such as power supplies, clock circuits, cache memory and the like as well as circuits that assist in executing the software routine stored in the memory 1440. As such, it is contemplated that some of the steps discussed herein as software processes may be implemented within hardware, for example as circuitry that cooperates with the processor 1430 to perform various steps. The controller 1400 also contains input/output (I/O) circuitry 1410 that forms an interface between the various functional elements communicating with the controller 1400. Although the controller 1400 of FIG. 2 is depicted as a generally purpose computer that is programmed to perform various control functions in accordance with the present invention, the invention can be implemented in hardware as, for example, an application specific integrated circuit (ASIC) or field programmable gate array (FPGA). As such, the process steps described herein are intended to be broadly interpreted as being equivalently performed by software, hardware or a combination thereof.
  • [0063]
    [0063]FIG. 13 depicts a high level block diagram providing a plurality of actions in accordance with an embodiment of the present invention. Specifically, the method 1300 of FIG. 13 is entered at step 1301 and, at step 1305, video server processes are partitioned. That is, at step 1305 a plurality of video server processes are segmented from each other such that individual processes may be executed using respective processing elements. At step 1310, each of a plurality of processing elements is associated with at least one respective process. That is, at step 1310, the video server processes partitioned at step 1305 are distributed across a plurality of processing elements.
  • [0064]
    At step 1315, priority is assigned to the various processing elements according to a hierarchy of video server processes. That is, at step 1315, the operating system assigns higher priority to those processing elements supporting video server processes that are high on a hierarchy of (or prioritized list of) video server processes. Such higher processes include memory access, stream transport encoding and the like. In general, higher order or higher priority video server processes are those that are directly responsible for insuring that the timing constraints discussed herein are met by the video server.
  • [0065]
    At step 1320, timing parameters, operating parameters and system loading parameters (and, optionally, other parameters) are monitored.
  • [0066]
    At step 1325, the hierarchy of video server processes is adjusted according to monitor timing parameters, changes in system loading or operating conditions and/or operating systems scheduler requirements. Thus, at step 1325, the relative priority level of video server processes is adjusted based upon various conditions within the video server and, optionally, its related transport network.
  • [0067]
    At step 1330, adjustments are made to the associations between video server processes and processing elements in response to processing element loading, monitored timing parameters, changes in system loading or operating conditions and/or operating system scheduler requirements. Thus, at step 1330, one or more of the plurality of processing elements may have an additional video server process associated with it or a video server process disassociated (i.e., removed).
  • [0068]
    The above-described method may be implemented using the systems and apparatus described below. Additional means of implementing the above method will be appreciated by those skilled in the art and informed by the teachings of the present invention.
  • [0069]
    The method 1300 of FIG. 13 is depicted as proceeding from step 1330 to step 1320 (i.e., repeating steps 1320-1330). It is noted that the hierarchy adjustments of step 1325 may result in changes of assigned priority to processing elements such as described above with respect to step 1315, and that association adjustments of step 1330 may result in association changes such as described above with respect to step 1310. Moreover, it is noted that in various embodiments of the invention either one of steps 1325 and 1330 are not utilized in the method 1300 of FIG. 13.
  • [0070]
    Advantageously, the invention enables the use of mass produced PC computing systems within the context of high accuracy streaming media applications.
  • [0071]
    While the foregoing is directed to embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.
Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US5675798 *May 16, 1996Oct 7, 1997International Business Machines CorporationSystem and method for selectively and contemporaneously monitoring processes in a multiprocessing server
US5778179 *Apr 7, 1997Jul 7, 1998Kabushiki Kaisha ToshibaSystem for flexible distributed processing and transaction processing suitable for nested transaction
US5799150 *Feb 20, 1997Aug 25, 1998Avid Technology, Inc.System for sending list of media data objects to server which may be read by client and receiving from the server indicator of allocated resource
US5987501 *Jul 14, 1998Nov 16, 1999Avid Technology, Inc.Multimedia system having server for retrieving media data as indicated in the list provided by a client computer
US6058414 *Jan 7, 1998May 2, 2000International Business Machines CorporationSystem and method for dynamic resource access in an asymmetric resource multiple processor computer system
US6212571 *Mar 4, 1997Apr 3, 2001Nec CorporationServer
US6223211 *Sep 9, 1999Apr 24, 2001Avid Technology, Inc.Apparatus and computer-implemented process for providing real-time multimedia data transport in a distributed computing system
US6711622 *Dec 31, 1997Mar 23, 2004Broadware Technologies, Inc.Video and audio streaming for multiple users
US6839764 *Jun 20, 2001Jan 4, 2005Intel CorporationDomain encapsulation
US7036123 *Apr 25, 2001Apr 25, 2006Sun Microsystems, Inc.System using fair-share scheduling technique to schedule processes within each processor set based on the number of shares assigned to each process group
US7222178 *May 31, 2002May 22, 2007Unisys CorporationTransaction-processing performance by preferentially reusing frequently used processes
US7227589 *Dec 22, 1999Jun 5, 2007Intel CorporationMethod and apparatus for video decoding on a multiprocessor system
US20020083468 *Feb 26, 2001Jun 27, 2002Dudkiewicz Gil GavrielSystem and method for generating metadata for segments of a video program
US20020093468 *Aug 29, 2001Jul 18, 2002Takayuki OuchiImage display apparatus
US20060233266 *Mar 3, 2006Oct 19, 2006Sharp Kabushiki KaishaContent processing device, content transmitting device, content receiving device, content distribution system, content synchronization program, and recording medium
Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US8161491Apr 27, 2010Apr 17, 2012Avaya Inc.Soft real-time load balancer
US8166485 *Aug 4, 2010Apr 24, 2012Avaya Inc.Dynamic techniques for optimizing soft real-time task performance in virtual machines
US8245234Jan 13, 2010Aug 14, 2012Avaya Inc.Credit scheduler for ordering the execution of tasks
US8484648 *Oct 19, 2009Jul 9, 2013International Business Machines CorporationHardware multi-threading co-scheduling for parallel processing systems
US8499303Apr 19, 2012Jul 30, 2013Avaya Inc.Dynamic techniques for optimizing soft real-time task performance in virtual machine
US8768104 *Jan 7, 2009Jul 1, 2014Pci Geomatics Enterprises Inc.High volume earth observation image processing
US20090232349 *Jan 7, 2009Sep 17, 2009Robert MosesHigh Volume Earth Observation Image Processing
US20110093638 *Oct 19, 2009Apr 21, 2011International Business Machines CorporationHardware multi-threading co-scheduling for parallel processing systems
U.S. Classification709/224, 348/E05.008
International ClassificationH04N5/00
Cooperative ClassificationH04N21/241
European ClassificationH04N21/241
Legal Events
Jan 26, 2004ASAssignment
Apr 28, 2004ASAssignment
Effective date: 20040421
Sep 20, 2004ASAssignment
Effective date: 20040824
Sep 24, 2008ASAssignment
Effective date: 20080913
Feb 13, 2012FPAYFee payment
Year of fee payment: 4